Terrence C. Stewart

Deconstructing and reconstructing ACT-R: Exploring the architectural space

Evaluating variations in the structure of computational models of cognition is as important as evaluating variations in the numerical parameters of such models. However, computational models tend not to be organized in such a way as to directly support such research. To address this need, we have taken the well-known cognitive architecture ACT-R, reduced it to its fundamental components, and reconstructed it. Our new system, Python ACT-R, facilitates exploration of the space of possible models and architectures based on the core ACT-R theory. The result has enabled us to examine the possibility of using basic ACT-R components such as the declarative memory system in new ways; for example, as the basis for a new visual attention system. Python ACT-R allows the same model definition syntax to be used to define both ACT-R models and new ACT-R components, as well as making explicit the processes specified by the ACT-R theory.

The Origin of Epistemic Structures and Proto-Representations

Organisms across species use the strategy of generating structures in their environment to lower cognitive complexity. Examples include pheromones, markers, color codes, etc. We provide a model of how such structures originate, and present a simulation where organisms with only reactive behavior learn, within their lifetime, to add such structures to their world to lower cognitive load. This implementation is then extended to show that the same underlying process could generate internal traces of the world (memories) in an internal environment. This model provides a novel account of the origin of internal representations. Further, as both external and internal traces are generated using the same mechanism, the model shows how an extended mind could be implemented. Also, as the stored internal traces develop entirely out of actions, these action components could be activated implicitly. This feature explains the origin of enactable and action-oriented mental content, suggested by recent experiments.

Multi-agent models of social dynamics in children

A series of computational models are presented which address the question of how peer relations change over time. We examine data from a standardized metric (CDC) that places school children in one of five categories: Popular, Rejected, Neglected, Controversial, and Average, and how such classifications change over time. A simple random model is shown to not match the empirical data, while a computational model and an ACT-R model are shown to match equally well, even though they are highly architecturally distinct. To test these models ability to give useful predictions in other domains, we introduced variation among the individuals in the models. For both models, we observed equivalent behavior that was consistent with the empirically known effects of Hostile Attribution Bias, variations in social skill, and shyness among others. This indicates that both models are capturing inherent underlying regularities of the social dynamics of peer relations in children. The relationship between these models and its implications are discussed.